OpenAI Model Autonomously Cracks 80-Year Math Problem, Shifting AI Research Stakes

OpenAI said that an internal general-purpose reasoning model autonomously solved the planar unit distance problem, a famous open problem in discrete geometry first posed by Paul Erdős in 1946, marking the first time one of its systems has cracked a long‑standing research question without step‑by‑step human guidance

The announcement sharpens an industry argument that frontier models are moving from assistant tools to original contributors in technical fields, with implications that stretch far beyond mathematics.

A New Bar for Autonomous AI Research

The company described the result as proof that advanced systems can hold a difficult argument together, combine ideas from distant areas of knowledge, and produce work that withstands expert review. External mathematicians verified the proof, which drew on tools from algebraic number theory.

OpenAI framed the milestone as part of a longer push toward more automated research. The lab said similar capabilities could one day support work in biology, physics, materials science, and medicine, where many problems are too large or complex for traditional teams to tackle alone.

Industry Race Heats Up

The breakthrough lands during a frantic stretch for the AI sector. OpenAI is reportedly preparing an IPO filing as soon as this week, just after a US jury cleared the company in a lawsuit brought by Elon Musk.

Rival Anthropic is on track for its first profitable quarter on projected revenue of $10.9 billion, while former OpenAI founding member Andrej Karpathy recently joined Anthropic to focus on frontier model research.

Labor and Strategy Questions Intensify

Autonomous problem-solving by AI is already reshaping how executives talk about high-skilled work. Citadel chief executive Ken Griffin recently warned that agentic AI is starting to replace PhD-level finance tasks in hours rather than months.

Some observers argue the next competitive edge in AI will not come from raw model quality but from access to real-world execution data that lets systems act, not just answer.

OpenAI said human judgment still anchors the work, with researchers choosing which problems matter and how to interpret results. What the new milestone changes is the range of problems a model can credibly take on alone.